People living with autoimmune diseases face an increased risk of cardiovascular disease, yet conventional risk calculators may not fully capture the complex inflammatory, clinical and treatment-related factors affecting this population. A recent article in Heart, Lung and Circulation highlights the potential for artificial intelligence—particularly deep learning—to improve cardiovascular risk assessment for these patients.
Traditional cardiovascular risk models generally rely on a limited set of variables, such as age, blood pressure, cholesterol, smoking and diabetes. For patients with conditions such as rheumatoid arthritis or systemic lupus erythematosus, these tools may overlook disease activity, chronic inflammation, medication exposure, imaging findings and other non-linear interactions that contribute to cardiovascular risk.
Deep-learning models offer a potential alternative by analysing large, multidimensional datasets and detecting patterns that may be difficult to identify using conventional statistical approaches. Relevant inputs could include electronic medical records, pathology results, medication histories, inflammatory biomarkers, imaging data and longitudinal changes in disease activity.
By integrating these diverse data sources, AI could support more personalised risk estimates, identify high-risk patients earlier and help clinicians target preventative measures such as lipid management, blood-pressure control and further cardiovascular investigation.
However, the article presents this as a promising prospect rather than an established clinical solution. AI models require validation in large and diverse autoimmune populations, transparent reporting, comparison with existing risk calculators and assessment of whether their predictions meaningfully improve patient outcomes. Poor-quality or unrepresentative training data could also produce biased or unreliable risk estimates.
The key message is that deep learning may help address an important gap in cardiovascular prevention for people with autoimmune disease. Its value will ultimately depend not only on predictive accuracy, but also on clinical interpretability, integration into routine care and rigorous prospective validation.
Read more: https://www.heartlungcirc.org/article/S1443-9506(25)01754-8/fulltext